THIS MONTH
a
POINTS OF VIEW
b
Pathways
npg
© 2016 Nature America, Inc. All rights reserved.
Apply visual grouping principles to add clarity to
information flow in pathway diagrams.
Pathway diagrams describe the connectivity and flow of information in biological systems. Remarkably similar representations can
depict everything from cell-signaling pathways to global ecological
networks.
Pathways are network diagrams in which molecules, cells or
species are represented by nodes and their relationships by edges.
Pathway diagrams benefit from strategies used to display networks1,
but additional requirements must also be met. First, they must clearly depict patterns in connectivity—the flow of information through
the pathway, encoded by the direction of the edges, is often the
primary purpose of the diagram. Second, both direct and indirect
relationships need to be clear for one to understand the pathway
as a whole. Encodings that remove indirect relationships, such as
adjacency matrices, are therefore inadequate.
One can use visual grouping to help create a hierarchy for the flow
of information in a pathway layout2 and clear alignment to emphasize
node relationships. Edges should connect to a fixed number of points
on node shapes (Fig. 1a). Basic arrowheads should be used; unnecessary stylizing or stretching of arrows should be avoided. Edge angles
should be limited to multiples of 30° or 45°, and curved edges can
be drawn easily using circular guides (Fig. 1b). We use 0.5-pt lines
for edges and equilateral-triangle arrowheads with sides 2.5 pt long3.
Conventionally, we expect information to flow left to right and top
to bottom. Diverging from this standard, as well as introducing asymmetry in the layout, can emphasize differences but should be done
sparingly and only when it adds to the reader’s understanding. Edges
that loop back to upstream nodes should flow clockwise (Fig. 1b).
Placing nodes on a grid assists eye movement across the figure
(Fig. 1c). Horizontal alignment of nodes emphasizes the flow of
information through a pathway, whereas radial alignment highlights source nodes (Fig. 1d). Local deviation from the grid pattern
may be necessary to avoid crossing of edges or collisions between
arrowheads (Fig. 1c).
a
Node
interface
b
Edge
curvature
c
Layout and routing grid
d
Square versus radial alignment
and curved versus diagonal edges
Figure 1 | Rectilinear and curved grids provide overall consistency.
(a) Fix the number and position of connector points and distribute edges
symmetrically in quadrants. (b) Create consistent edge curvature with
grids based on circles and rounded rectangles. (c) Nodes positioned on a
square grid with circular edge curvature. Connecting adjacent edges with
overlapping arrowheads can be locally adjusted from their circular path
(dotted line) by vertical stretching (orange) or shrinking (blue) of the
edges. (d) Neighbors (gray) of a source node (dark gray) can be aligned
horizontally (left) or radially (right). Arrow length can be altered in radial
alignment to create groups. Curved (blue) and diagonal (orange) edges are
locally adjusted from a 45° guide to avoid overlap between arrowheads.
Figure 2 | Creating a clear pathway diagram using a grid and directional
information flow. (a) Example of a pathway diagram with unorganized
information flow, unnecessary visual detail (e.g., membrane lipids) and
no visual continuity along pathways. The dashed line represents the cell
nucleus. (b) Pathway from a with redundant visual encodings removed and
main points emphasized by visual grouping. Color and shape variations have
been removed except for those highlighting a molecule of interest (orange),
the products of the pathway (blue) and a membrane protein complex
(green). Gray lines are layout guides and would not be included in the final
figure; the solid lines provide a grid, and the dotted lines highlight the
strong grouping effect of visual connection (circular paths) and proximity.
Strong relationships between pathway components can be illustrated using connection and enclosure4. Edges act to group nodes
via connection, whereas enclosure can be used to group nodes in
shared compartments, such as the nucleus (Fig. 2).
Associating nodes through similarity (e.g., color or shape) or
proximity can highlight parts of a pathway without interrupting
the groupings created by connection and enclosure. For similarity
groups to be effective, unnecessary variation in the color or shape
of nodes should be avoided, except when used to highlight nodes
(Fig. 2b). Proximity grouping can be achieved with negative space—
an empty row or column on the grid around the group adds visual
emphasis (Fig. 2b). Differences can be identified with labels or with
an unambiguous shape associated with a specific protein class. For
example, in Figure 2b we show the seven transmembrane domains
of the GPCR, and the G protein as a green complex.
The use of grouping in pathway diagrams can provide alternate
visual entry points. In a busy pathway, it can be difficult to work
from start to finish through all possible paths. When important
node subtypes are easily identifiable, a pathway diagram can be
examined from several directions instead of strictly serially.
Adding labels to nodes is often a challenge—names of genes
and protein complexes can be long, but altering node shapes to fit
labels dilutes grouping effects (Fig. 2a). It is a good idea to choose
node shapes that accommodate the longest label, or abbreviate
names. Keep node colors desaturated to avoid loss of contrast with
the text, and avoid visual garnishes such as gradients and drop
shadows.
Next month we will explore how to encode multiple variables
in pathway diagrams by taking a closer look at neural circuits.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Barbara J. Hunnicutt & Martin Krzywinski
1. Gehlenborg, N. & Wong, B. Nat. Methods 9, 115 (2012).
2. Wong, B. Nat. Methods 7, 941 (2010).
3. Wong, B. Nat. Methods 8, 701 (2011).
4. Wong, B. Nat. Methods 8, 101 (2011).
Barbara J. Hunnicutt is a research assistant at Oregon Health and Science
University. Martin Krzywinski is a staff scientist at Canada’s Michael Smith
Genome Sciences Centre.
NATURE METHODS | VOL.13 NO.1 | JANUARY 2016 | 5